HAL Id: hal-01913876https://hal.inria.fr/hal-01913876
Submitted on 6 Nov 2018
HAL is a multi-disciplinary open accessarchive for the deposit and dissemination of sci-entific research documents, whether they are pub-lished or not. The documents may come fromteaching and research institutions in France orabroad, or from public or private research centers.
L’archive ouverte pluridisciplinaire HAL, estdestinée au dépôt et à la diffusion de documentsscientifiques de niveau recherche, publiés ou non,émanant des établissements d’enseignement et derecherche français ou étrangers, des laboratoirespublics ou privés.
Distributed under a Creative Commons Attribution| 4.0 International License
Optimal Scheduling of Multiproduct Pipeline SystemUsing MILP Continuous Approach
Wassila Abdellaoui, Asma Berrichi, Djamel Bennacer, Fouad Maliki, LatéfaGhomri
To cite this version:Wassila Abdellaoui, Asma Berrichi, Djamel Bennacer, Fouad Maliki, Latéfa Ghomri. Optimal Schedul-ing of Multiproduct Pipeline System Using MILP Continuous Approach. 6th IFIP InternationalConference on Computational Intelligence and Its Applications (CIIA), May 2018, Oran, Algeria.pp.411-420, �10.1007/978-3-319-89743-1_36�. �hal-01913876�
Optimal scheduling of multiproduct pipeline
system using MILP continuous approach
Abdellaoui Wassila1*, Berrichi Asma1, Bennacer Djamel2, Malik Fouad
1,
Ghomri Latefa1
1 University Abou Bekr Blkaid, Dept. of Electrical Engineering Tlemcen, Algeria 2 University Abou Bekr Blkaid: Dept. of Mechanical Engineering Tlemcen, Algeria
Abstract. To date, the multiproduct pipeline transportation mode has nationally and
internationally considerably evolved thanks to his efficiently and effectively of transporting
several products. In this paper, we focus our study on the scheduling of a multiproduct
pipeline system that receives a number of petroleum products (fuels) from a single refinery
source in order to be distributed to several storage and distribution centers (depots). Mixed
Integer Linear Programming (MILP) continuous mathematical approach is presented to
solve this problem. The sequence of injected products in the same pipeline should be
carefully studied, in order to meet market demands and ensure storage autonomy of the
marketable pure products in the fuels depots on the one hand and to minimize the number
of interfaces; Birth zone of mixture between two products in contact and in sequential flow,
which may hinder the continuous operation of the pipeline system, by the necessity of
additional storage capacity for this last mixture, that is in no way marketable and requires
special processing operations. This work is applied on a real case of a multiproduct pipeline
that feeds the western and southwestern region of Algeria with fuels. The obtained results
based on the MILP continuous approach give an optimal scheduling of the multiproduct
transport system with a minimized number of interfaces.
Keywords: multiproduct pipeline; Scheduling; fuels; MILP continuous approach.
1 Introduction
Several million barrels of crude oil and oil derivatives such as diesel, gasoline, kero-
sene, jet fuel and liquid petroleum gases (LPG) are moved daily around the world in
imports and exports. These products can be transported with different modes: roads,
railroads, vessels, and pipelines among which the latter is the safest and least expen-
sive way to distribute large quantities of energy products from source generally repre-
sented by the refineries to distribution terminals located in the different place [1, 2].
Among the best benefits of the pipeline, is its ability to operate 24 hours a day, re-
gardless of weather conditions, with much lower operating costs than other inland
transport modes [3].
Pipelines were first utilized by oil transportation companies of crude petroleum
and its derivatives, where demand for these petroleum products is high. Oil industries
are decided to expand pipeline utilization due to its low operating cost [4].
Products are injected in the pipeline one after the other in batch, in order to be
transported to several distribution centers. Generally, we can distinguish two forms of
the pipeline; the first one is straight-structured pipeline circulation of batches inside
the pipeline, where the flow is established in one direction from source to centers. The
second case is where the pipeline transfer products in two directions, called bidirec-
tional pipelines, the last is called tree-structured pipeline because it takes the tree form
and at each tree branch called "segment", a depot is located.
In this research, we are interested in the scheduling of multiproduct pipeline take
the form of a straight-structured pipeline .Our goal is to find the optimal sequence of
the new batches injection inside the pipeline; What allows to satisfy in wanted time
daily requests of terminals and to minimize contaminant interface which results be-
tween the different batch of product and ensures an autonomy of stock of 20%
2 Literature Review
Several works on the multiproduct pipeline systems problems have appeared over the
last years. Many authors have presented different approaches for scheduling multi-
product pipeline systems in the literature: knowledge-based search techniques and
mathematical programming approaches such as Mixed Integer Linear Programming
(MILP) used by [5-9] formulations or Nonlinear Mixed Integer Programming
(MINLP) formulations. The last can be divided into two approaches: discrete MILP
approach and continuous MILP approach [10].
(Hane and Ratliff 1995), presented a discrete MILP model to transport several
products from the refinery to diverse depots; this problem is divided into sub-
problems solved by branch and bound method [11].
(Rejowski and Pinto 2003), were studied a multiproduct system connected a
unique refinery to several distribution centers that assure the demand of the local con-
sumer markets. They proposed MILP model based on discrete time for scheduling this
system .The model assumed in the beginning that le pipeline is divided into segments
and each segment is divided into packs of equal volume, in the second they eliminated
this assumption which says that the packs have the same volume [12].
The result of objective function ensures minimization of significant most opera-
tional costs, such inventory costs at the refinery and distribution centers, pumping
costs and interface costs between adjacent products inside the pipeline, and moreover,
the optimal sequence of injected products in the origin of pipeline and also the inven-
tory levels at depot and refinery [1].
A year later, the same author adds special and non-intuitive practical constraints to
the original model that can minimize volume between adjacent products (contami-
nant) inside the pipeline, and after, they included at the first MILP model a set of
integer cuts that are based on the demand of depot and pipeline segment initial inven-
tories [13].
(Cafaro and Cerda 2004), propose to study the problem treated by [12] with a nov-
el MILP continuous mathematical formulation, which does not require division of
pipeline into packs and time discretization. So, it has continuous representation in
time and volume. Something special for this presentation, that allows minimizing the
number of binary variables and gives the best sequence of injected products into the
pipeline with the best costs [14].
(Relvas et al. 2006) , decided to integrate the distribution center operation with the
multiproduct pipeline operation in their study. Their objective is to do both, the
scheduling of the multi-product pipeline transport and the supply management in the
depots. They have applied this model to a real case of a Portuguese oil distribution
company[15]. (Cafaro and Cerda 2008b), studied the problem of scheduling of multi-
product with different assumptions where clients demand was dynamic and they used
a multi period moving horizon. They were giving good results appropriate to the real
case with very short computational time [16].
(Relvas et al. 2006), developed a heuristic that can find the optimal sequence of
pumped products into the pipeline, they use this heuristic for a short to medium term
horizon. (MirHassani et al. 2011), presented an algorithm for the long-term planning
of a simple multiproduct pipeline. The algorithm used a continuous MILP model for a
short-term schedule to come to a long-term schedule. The objective function was
minimizing the penalty costs of non-use of pipeline capacity, interface costs and cost
of no satisfied customer demand[17].
(Rejowski and Pinto 2008),The authors present a new mixed-integer non-linear
programming (MILP) for intermittent multiproduct pipeline scheduling which takes
into account several constraints like different flow rates, pipeline works intermittently.. etc
This representation gives a good result comparing with (Rejowski & Pinto, 2004) which use a
discrete representation of time [18].
Recently, the authors are interested in pipeline networks with multiple origins and
destinations have been studied; Like what they did Cafaro DC, Cerda J (2016), Intro-
duce a new tool for optimizing the short-term planning of petroleum product pipe-
lines; Mixed-integer linear programming (MILP) model is expanded to treat pipeline
networks with multiple sources, unidirectional flow and a single pipeline between
every pair of the adjacent distribution center [19].
3 Problem Statement
In this paper we aim to study the activities scheduling in multiproduct pipeline system
of a straight-structured unidirectional pipeline type, connecting a unique origin (refin-
ery) to multiple distribution centers (in the case study we have two centers.
The experiment site is a 168-km multiproduct pipeline linking a refinery in western
Algeria to storage and fuel distribution centers. Fuels moved in batches (We note four
pure fuel batches, P1, P2, P3 and P4) from the refinery tank farm through pumping
station without any physical separation between adjacent products. An area of mixture
was established between batches, where this last zone of the mixture progress until
reaching the terminal at pipeline’s end. The number of mixture depends on the num-
ber of initial products injected in the pipeline [20]. Figure 1 shows the physical struc-
ture of studied pipeline system.
The problem purpose is to determine sequence and volumes of new product batches to
be pumped in the pipeline, in order to meet market demands and ensure products stor-
age autonomy or the security stock in depots (fixed to 20% of overall storage capacity
of each center) with number of interface between adjacent products p and p’ inside
the pipeline minimized (Reduced).
Fig.1.Single unidirectional multiproduct pipeline system
a. 3.1 Model Assumptions
(a) All products move in the pipeline without any physical separation between every
two products in contact.
(b) The pipeline is always full, so if we think to receive a quantity of products from
all the depots, it’s necessary to inject the same amount at the origin of the pipe-
line.
(c) At any new pumping operation, only a unique product i.e. single batch is injected
into the pipeline.
(d) Length of the planning horizon is fixe.
(e) Volume between adjacent products in pipeline (contaminant) was fixe.
(f) The scheduling model will meet the demand of products by the depots for daily
sales to satisfy customer.
4 Optimization Model
Nomenclature
Sets
I Set of batch (Iold ∪ new)
Iold Set of batch inside pipeline
Inew Set of new batch that will be injected in pipeline
J Set of distribution center
P Set of Product
Parameters
Li Time of injection of batch i Si Time starting pumping the batch i mix(p, p’)
Volume interface between the batch i and batch i + 1 include le
product p after p’
Variables
vmh,j,i Volume batch h delivered distribution center j from the pipeline
during the injection of the batch i nvsp,j,i Stock level of product p in the distribution center j at the end of the
batch i injection
vmh,j,i Volume batch h delivered distribution center j from the pipeline
during the injection of the batch i INVi,p,p’ Interface volume between batch i and (i − 1) if they contain prod-
ucts p and p’
yi,p Binary variable denoting that product p is contained in batch i whenevery i, p = 1 Otherwise yi,p = 0
4.1 Objective Function
Problem objective Function is given in equation (1) consisted to minimize the inter-
face volume between two adjacent products in the pipeline.
𝑚𝑖𝑛 ∑ ∑ ∑ 𝐼𝑁𝑉𝑖,𝑝,𝑝’
𝑝’∈𝑃
(1)
𝑝∈𝑃𝑖∈𝐼
4.2 Constraints
4.3.1 Product Allocation to Batch
yi,p is the binary variable, shows that product 𝑝 is contained in batch 1 it takes value
yi,p=1 , if else yi,p = 0. And every batch can, at more, take one product so:
∑ 𝑦𝑖,𝑝
𝑝∈𝑃
≤ 1 ∀ 𝑖 ∈ 𝐼𝑛𝑒𝑤 (2)
The new batch i will be injected after batch i − 1 so if batch i − 1 take any product
∑ yi−1,pp∈P
= 0 the batch i was not injected.
∑ 𝑦𝑖,𝑝
𝑝∈𝑃
≤ ∑ 𝑦𝑖−1,𝑝
𝑝∈𝑃
∀ 𝑖 ∈ 𝐼𝑛𝑒𝑤 (3)
4.3.2 Batch sequencing
The injection of a new batch 𝐢 ∈ 𝐈𝐧𝐞𝐰 in the pipeline should start after the end of injected batch 𝐢 − 𝟏.
𝑺𝒊 ≥ 𝑺𝒊−𝟏 + 𝑳𝒊−𝟏 ∀𝒊 ∈ 𝑰𝒏𝒐𝒖𝒗𝒆𝒂𝒖𝒏𝒆𝒘(𝒊 ≥ 𝟐) (4)
4.3.3 Interface Volume Between Two Successive
Inside the multiproduct pipeline, there is no physical separation between different products, so we record certain volume of intermixing between the two adjacent batches which is assumed a constant value and it is presented with 𝐦𝐢𝐱(𝐩, 𝐩’). The continuous variable 𝐈𝐍𝐕𝐢,𝐩,𝐩’ that presents interface volume between batches 𝐢 and 𝐢 + 𝟏 take the value of 𝐦𝐢𝐱(𝐩, 𝐩’), if product 𝐩 was located in batch 𝐢 and product 𝐩’ was located in batch i+1.
∑ ∑ ∑ 𝐼𝑁𝑉𝑖,𝑝,𝑝’ ≥ 𝑚𝑖𝑥(𝑝, 𝑝’) ∗ (𝑦𝑖,p + 𝑦𝑖+1,p’ − 1)
𝑝’∈𝑃𝑝∈𝑃𝑖∈𝐼
∀ i ∈ I, i < |I|, p, p’ ∈ P (5)
4.3.4 Inventory in distribution center
The inventory of products in distribution center j at the end of injection batch i was
equal to the inventory product at the end of pumped batch i − 1 nvsp,j,i−1 by adding
sum of product volume transferred to the distribution center during pumped batch 𝑖 (vmph,p,j,i for depot 1 and vsqh,p,j,i for depot 2 ) minus quantity delivery to clients.
𝑛𝑣𝑠𝑝,𝑗,𝑖 = 𝑣𝑖𝑛𝑡𝑝,𝑗 + ∑ 𝑣𝑚𝑝ℎ,𝑝,𝑗,𝑖ℎ∈𝐼,ℎ≤𝑖 − 𝑣𝑜𝑚𝑝,𝑗,𝑖 (6)
∀𝑝 ∈ 𝑃, 𝑗 ∈ 𝐽(𝑗 < |𝐽|), 𝑖 ∈ 𝐼𝑛𝑜𝑢𝑣𝑒𝑎𝑢(𝑖 = 𝑓𝑖𝑟𝑠𝑡 𝑛𝑒𝑤 𝑏𝑎𝑡𝑐ℎ)
𝑛𝑣𝑠𝑝,𝑗,𝑖 = 𝑛𝑣𝑠𝑝,𝑗,𝑖−1 + ∑ 𝑣𝑚𝑝ℎ,𝑝,𝑗,𝑖ℎ∈𝐼,ℎ≤𝑖 − 𝑣𝑜𝑚𝑝,𝑗,𝑖 (7)
∀𝑝 ∈ 𝑃, 𝑗 ∈ 𝐽(𝑗 < |𝐽|), 𝑖 ∈ 𝐼𝑛𝑒𝑤
To ensure that the level inventory in distribution center was grater than or equal to the
stock of security, we use the following constraint, where ssp,j presents stock of
security that is fixed at 20% of the overall stock capacities of each dept.
𝑛𝑣𝑠𝑝,𝑗,𝑖 ≥ 𝑠𝑠𝑝,𝑗 ∀𝑝 ∈ 𝑃, 𝑗 ∈ 𝐽, 𝑖 ∈ 𝐼𝑛𝑒𝑤 (8)
5 Result and Discussion
5.1 Given : These data is harvest according to our real case study
Table 1. Daily demand of depots
product Daily demand [𝒎𝟑]
Depot 1 Depot2
P1 1200 3000
P2 400 800
P3 80 150
P4 - 150
Table 2.Tanks storage capacities, products inventories and products security inventories
Level Inventory [𝑚3]
product Depot 1 Depot 2
Capacity Initial inven-
tory
security inven-
tory
Capacity Initial inven-
tory
security inven-
tory
P1 6000 814 1200 22000 5572.4 4400
P2 1700 809 340 9500 3394.6 1900
P3 450 196 90 1000 996 200
P4 - - - 5000 3284.7 1000
Table 3.Interface volume and possible sequences between subsequent products inside the
pipeline, p and 𝑝′
product Volume Interface [𝑚3]
P1 P2 P3 P4
P1 28 30
P2 28 0
P3 30
P4 30
Table 4.Initial batches volume inside pipeline at t = 0
Initial batchs inside pipeline
Volume [𝒎𝟑]
Batch01 (p2) 520
Batch01 (p1) 9580
5.2 Analyses Result and Discussion
The value of function objective was 320 m3 presented the total of interface mixture
result between adjacent product inside the pipeline. They presented 2.30% of total
volume, so we can say that the result was good more than can be satisfied the daily
demand of client. The optimal sequence of the new batch injected inside the pipeline
that can minimize the objective function was:
𝒑𝟑(𝟏) − 𝒑𝟏(𝟐) − 𝒑𝟐(𝟑) − 𝒑𝟏(𝟒) − 𝒑𝟑(𝟓) − 𝒑𝟏(𝟔) − 𝒑𝟐(𝟕) − 𝒑𝟏(𝟖)
This sequence can minimize the number of an interface between adjacent product p
and p’ and therefore minimize the objective function by satisfying the sales forecasts
of centers and finally customer demands.
Table5. In Volume of a new batch injected inside the pipeline
N° of batch Volume of batch [𝑚3]
𝑝1 𝑝2 𝑝3 𝑝4
Batch 1 386
Batch 2 4776.5
Batch 3 8298.4
Batch 4 15000
Batch 5 1029.1
Batch 6 13710
Batch 7 10000
Batch 8 2280
Table5 shows the volumes of a new batch injected inside the pipeline satisfying the
centers demands (depot 1 and 2) and ensure level tanks higher or equal than the stock
of security.
Table6. Level tanks in the depot 1 at the end of injection of the new batch
Batch Tank 𝒑𝟏 Tank 𝒑𝟐 Tank 𝒑𝟑
Volume
[𝑚3]
% Volume
[𝑚3]
% Volume
[𝑚3]
%
Initial 814 13.6 809 47.6 196 43.6
Batch 1 1200 20 809 24,1 116 25.78
Batch 2 2414.9 40.24 409 24.06 116 25.78
Batch 3 4800 80 400 23.53 347.5 83.23
Batch 4 2400 40 1480 87.06 187.55 41.67
Batch 5 1200 20 1080 63.53 107.55 23.9
Batch 6 4256 70.93 680 40 426.67 94.81
Batch 7 3200 53.33 340 20 213.33 47.41
Batch 8 1600 26.67 1260 74.12 106.67 23.71
Table 7 Level tanks in the depot 2 at the end of injection of the new batch
Batch Tank 𝒑𝟏 Tank 𝒑𝟐 tank 𝒑𝟑 Tank 𝒑𝟑
Volume
[𝑚3] % Volume
[𝑚3] % Volume
[𝑚3] % Volume
[𝑚3] %
Initial 5572,4 25,3 3394,6 35,7 996 99,6 3284,7 65,7
Batch 1 4400 20 2594.6 27.31 1000 90,6 3284.7 62,6
Batch 2 4400 20 3100.6 32.64 905.55 90,6 3134.7 53,7
Batch 3 5410.7 24.59
2300.6 24.22 755.55 90,6 2984.7 53,7
Batch 4 6370.9 28.96
6700 70.53 500 45,6 2684.7 53,7
Batch 5 4400 20 5900 62.11 350 100 2534.7 53,7
Batch 6 10455 47.52 5100 53.68 200 55 2384,7 47,7
Batch 7 11402
51.83 3800 40 400 55 2384.7 38,7
Batch 8 5750.9 2 6.14 1900 20 200 40 1784,7 35,7
In the table 6 and 7 we can see that tanks of all product at each depot (depot 1 and
depot 2) at the end injection of the new batch; The level inventory was more than the
stock of security so we assure that the probability of no satisfied the demands of cli-
ents will be decrease so always the demands of distribution centers was satisfied
6 Conclusions
Scheduling of a unidirectional multiproduct pipeline connecting a single refined to
multiple distribution terminals is carried out with The MILP continuous approach.
The MILP continuous can minimize the number of variable discussion compared to
the MILP discrete and gives best results in the multiproduct pipeline problem sched-
uling case.
The MILP continuous was applied on the Algerian multiproduct pipeline scheduling.
The pipeline in question links a refinery in western Algeria to two storage and fuel
distribution centers. It has be seen that the use of MILP continuous approach gives the
best result by taking into account the different operating conditions of the multiprod-
uct pipeline which is served as experiment site.
The obtained results present an optimal solution that gives the sequence of new batch-
es to be introduced in the pipeline. The flow configuration has reduced the number of
interfaces between products p and p’ in contact and in a sequential flow, unlike to the
current planning system of the company which is not based on recognized models.
Furthermore, we were able to quietly ensure autonomy storage that exceeds the re-
quired safety stock set at 20%.
Finally, our work based on MILP continuous approach can be considered as a deci-
sion support tool, to be called for use in the case of planning and scheduling of multi-
product pipeline use the MILP continuous approach for solve this type of problems.
References
1. Mostafaei, H., Alireza, GH.: A general modeling framework for the long-term scheduling of
multiproduct pipelines with delivery constraints. Ind. Eng. Chem. Res 53 (17), 7029–7042
(2014).
2. LNCS Homepage, http:// www.aopl.org , last accessed 2001/12/12.
3. Cafaro, V., Cafaro, D., Mendez, C.A., Cerdá, J.: detailed scheduling of oil products pipelines
with prallel batch input at intermediate source. Chemical Engineering Transactions 32, 1345-
1350 (2013).
4. Hobson, GD., Pohl, W.: Modern Petroleum Technology. 5th edn. John Wiley & Sons, Eng-
land (1982).
5. Cafaro, V.G., Cafaro, D.C., Méndez, C.A., Cerdá, j.: Detailed Scheduling of Operations in
Single-Source Refined Products Pipelines. nd. Eng. Chem. Res,50 (10), 6240–6259(2011).
6. Cafaro, V.G., Cafaro, D.C., Méndez, C.A., Cerdá, j.:Detailed Scheduling of Single-Source
Pipelines with Simultaneous Deliveries to Multiple Offtake Stations. Eng. Chem.
Res, 51 (17), 6145–616 (2012).
7. Mostafaei, H., Castro, PM. Ghaffari-Hadigheh, A.: A Novel Monolithic MILP Framework
for Lot-Sizing and Scheduling of Multiproduct Treelike Pipeline Networks. Eng. Chem.
Res. 54 (37), 9202–9221 (2015)
8. Zaghian, A, Mostafaei, H. :An MILP model for scheduling the operation of a refined petrole-
um products distribution system Ali Zaghian. Operational Research - An International Jour-
nal(ORIJ) .16,513–542(2016).
9. Castro, PM., Mostafaei, H.: New Continuous-Time Scheduling Formulation for Multiproduct
Pipelines.40,1381-1386(2017).
10. Cafaro, DC., Cerdá, J.: Efficient Tool for the Scheduling of Multiproduct Pipelines and Ter-
minal Operations. Ind. Eng. Chem. Res 47, 9941–9956(2008).
11. Hane, C.A., Ratliff, H.D.: Sequencing inputs to multi-commodity pipelines. Annals of Opera-
tions Research 57,73-101(1995).
12. Rejowski, R., Pinto, J.M.: Scheduling of a multiproduct pipeline system. Comput Chem Eng
27,1229–1246 (2003).
13. Rejowski, R., Pinto, J.M: Scheduling of a multiproduct pipeline system. Comput Chem Eng
28, 1511–1528 (2004).
14. Cafaro, D.C., Cerda, J.: Optimal scheduling of multiproduct pipeline systems using a non-
discrete MILP formulation. Comput Chem Eng 28,2053–2068(2004).
15. Relvas, S., Matos, H.A., Barbosa-Povoa, A..P.F.D., Fialho, J., Pinheiro, A.S.:Pipeline sched-
uling and inventory management of a multiproduct distribution oil system.. Ind Eng Chem
Res 45,7841–7855(2006).
16. Cafaro, D.C., Cerda, J.: Dynamic scheduling of multiproduct pipelines with multiple delivery
due dates. Comput Chem Eng 32,728–753 (2008b).
17. MirHassani, S. A., Moradi, S.,Taghinezhad, N.:Algorithm for long-term scheduling of mui
product pipelines. Eng. Chem. Res50 (24), 13899–13910 . (2011).
18. Rejowski, R., Pinto, J.M.: A novel continuous time representation for the scheduling of pipe-
line systems with pumping yield rate constraints. Comput Chem Eng 32,1042–1066(2008).
19. Cafaro, DC., Cerdá, J: Short-term operational planning of refined products pipelines. Optimi-
zation and Engineering 18, 241–268(2016).